Camera pose estimation based on 2D image and 3D point cloud fusion

Table 1. Detail of labcore dataset
View table
View in Article
Table 1. Detail of labcore dataset
数据集名称 Labcore
RGB 640×480
Depth 640×480
深度值（16位）单位 mm
帧数 4 000
帧率 30 fps
训练样本数 2 000
测试样本数 2 000

Table 2. Performance comparison of camera pose estimation algorithms

View table

View in Article

Table 2. Performance comparison of camera pose estimation algorithms

数据集场景	数据集组成 Train Test	位姿估计方法
数据集场景	数据集组成 Train Test	PoseNet^［5］	DSAC++^［15］	SANet^［20］	Pixloc^［16］	多阶段场景回归（本文）
Chess	4 000 2 000	32.0 cm/4.06°	2.0 cm/0.5°	3.0 cm/0.88°	2.0 cm/0.80°	1.8 cm/0.64°
Fire	2 000 2 000	47.0 cm/7.33°	2.0 cm/0.9°	3.0 cm/1.08°	2.0 cm/0.73°	1.8 cm/0.86°
Heads	1 000 1 000	29.0 cm/6.00°	1.0 cm/0.8°	2.0 cm/1.48°	1.0 cm/0.82°	1.2 cm/0.72°
Office	6 000 4 000	48.0 cm/3.84°	3.0 cm/0.7°	3.0 cm/1.00°	3.0 cm/0.82°	2.7 cm/0.81°
Pumpkin	4 000 2 000	47.0 cm/4.21°	4.0 cm/1.1°	5.0 cm/1.32°	4.0 cm/1.21°	4.0 cm/1.11°
RedKitchen	7 000 5 000	59.0 cm/4.32°	4.0 cm/1.1°	4.0 cm/1.40°	3.0 cm/1.20°	4.0 cm/1.10°
Stairs	2 000 1 000	47.0 cm/6.93°	9.0 cm/2.6°	16.0 cm/4.59°	5.0 cm/1.30°	3.5 cm/1.01°

Table 3. Time consumption comparison of camera pose estimation algorithms
View table
View in Article
Table 3. Time consumption comparison of camera pose estimation algorithms
处理步骤 DSM^［9］ DSAC++^［15］ Pixloc^［16］本文
图像检索（单帧） 0.17 — 0.17 0.17
位姿估计（单帧） 0.21 0.2 0.68 0.18
总耗时（单帧） 0.38 0.2 0.85 0.35

Tools

Get Citation

Copy Citation Text

Jia-le ZHOU, Bing ZHU, Zhi-lu WU. Camera pose estimation based on 2D image and 3D point cloud fusion[J]. Optics and Precision Engineering, 2022, 30(22): 2901

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites